Measurement Accuracy: Real Or Imagined

Scott Swinford
Vice President of Engineering
Pacific Instruments, Inc.
Concord, CA
IMAC XXVII
February 2009
Orlando, FL
Measurement Accuracy: Real Or Imagined
Quality Products
Quality Products Come From Quality Design & Testing.
Quality Testing Requires Measurement Accuracy.
Measurement Accuracy Comes From:
Decreasing Error Sources
Increasing Measurement Resolution
Proper Calibration
Understanding And Managing All Elements
Of The Measurement
Measurement Accuracy: Real Or Imagined
Elements Of A Measurement
Transducer You
Simply Stated: A Measurement Is Made Using Very Few Elements.
The Elements Can Be Complex And They Do Affect Each Other.
The Best Measurement That Can Be Made Depends On The
Transducer Accuracy.
After That, The Best Measurement That Can Be Made Depends On The
Measurement System Accuracy.
After That, The Best Measurement That Can Be Made Depends On
What You Do (Or Fail To Do) Properly.
If All Goes Well A Quality Measurement Is Made.
Measurement
System
Measurement Accuracy: Real Or Imagined
Specification Sheets: A Good Starting Point
Specification Sheets Provide Accuracy Information.
Each Element Involved In The Measurement Needs
To Be Considered.
Beware Of Specsmanship Issues.
Transducer
Specifications
Measurement System
Specifications
Calibration Standard
Specifications
Measurement Accuracy: Real Or Imagined
Specsmanship: What Is It?
Describing An Elements Accuracy Involves Providing Information
About All Possible Error Sources.
There Is No Standard Method Of Reporting This Information.
Specsmanship Is The Process Of Reporting The Information In A
Fashion Where It Looks Best To The Reader.
This Can Involve Omitting Critical Details Or Information About
How The Error Information Should Be Applied.
It Is Not Always A Trick To Deceive You. It Can Be Due To:
Not Wanting Competitors To Know About The Product.
Needing To Have More Information About Your Usage Of The
Product To Properly Specify The Error.
When In Doubt. Ask About.
Measurement Accuracy: Real Or Imagined
System 1 Common Mode Rejection Ratio Spec.
60 dB + Gain In dB To 120 dB For Balanced Input And
110 dB For A 350 Ohm Source Unbalanced,
DC to 60 Hz.
Specsmanship Example
System 2 Common Mode Rejection Ratio Spec.
130 dB
Common Mode Rejection Ratio (CMRR) Depends On Common Mode
Signal Frequency, Amplifier Gain, And Input Lead Balance.
On The Surface, System 2 Would Seem Better.
However, The Specification Indicates Nothing About How The Product
Will Be Used In Your Facility.
System 2s Spec Should Be Further Researched.
130 dB Is Most Likely Not The Level Of CMRR You Will Experience.
Measurement Accuracy: Real Or Imagined
E
Total
= (E
N
2
)
When The True Error Specifications Are Known
The Root Sum Square (RSS) Can Be Used To Calculate
The Total Measurement Error Expected.
Where
E
Total
Is The Total Measurement Error.
E
N
Is The Error Associated With
Element N of The Measurement.
Total Measurement Error
Example:
Transducer Accuracy +/- 1.0%
Measurement System Accuracy +/- 0.5%
E
Total
= SQRT((0.01)
2
+ (0.005)
2
)
E
Total
= 0.012 = 1.2%
Higher Than
Either Accuracy
(Not The Sum)
Measurement Accuracy: Real Or Imagined
Normalization & Site Specifics
One Task You Must Not Overlook Is Error Source Normalization.
Since Errors Are Treated As Percentages You Need To Ask
Percentage of What?
Percentage of Full Scale Range Is Typically Used In Specifications.
However, Vendor To Vendor Full Scale Range Can Vary.
Errors That Vary Due To Site Specifics (i.e. Temperature) Need To
Take Into Account The Intended Environment.
Remember To Compare Apples To Apples.
Measurement Accuracy: Real Or Imagined
Transducer Element Error
Transducer
Transducer Errors Depend On:
Transducer Type.
Measurement Environment.
Transducer Installation & Handling.
Transducer Error Defines The Best Measurement
That Can Be Made With The Transducer.
Measurement Accuracy: Real Or Imagined
Pressure Transducer Error Example
This Pressure Transducers Typical Measurement Error Is 0.173% of Full Scale.
The Value Accounts For 1g Of Acceleration And 10 F Of Temperature Variation.
Assuming The Temperature Variation And Acceleration Are Zero
The Typical Measurement Error Would Be 0.1 %.
The Environmental Factors Almost Double The Error.
This Error Can Not Be Calibrated Out.
Site
Specifics
Measurement Accuracy: Real Or Imagined
Measurement System Errors
Signal
Conditioning
Filtering
Digitization Amplification
EU
Calibration
Standards
Voltage
Calibration
Standards
Excitation
Power
Cabling
The Measurement System
Includes Many Parts.
Each Part Can Contribute To The
Measurement System Error.
Some Parts Have Multiple
Settings And Their Errors Will
Depend On The Setting Used.
Some Parts Will Not Be Used
In Some Measurements.
The Total Measurement System
Error Is The RSS Of The Errors
For The Parts Used At The
Settings Used.
Some Errors Will Be Nullified
Using Calibration Standards.
Measurement
System
Measurement Accuracy: Real Or Imagined
Which Errors Must Be Considered?
Data Conversion Method
Voltage Measured Indicates Engineering Unit (EU) Value.
Thermocouple Lookup
Transducer Datasheet
Must Ensure Voltage Measured Is Calibrated
Must Take Into Account Cable Losses
Must Ensure Excitation At Transducer Is Correct
In Situ (End To End) Calibration
Shunt Calibration
Controlled & Variable Environment
Must Ensure Calibration Stimuli
Transducer Type
Is Signal Conditioning Used (e.g. Bridge Completion)?
Is Excitation Used?
Measurement Accuracy: Real Or Imagined
Excitation Error
Excitation Errors Affect Transducers That Require Excitation Voltage Or Current.
The Level Error Specification Can Be Ignored If End To End EU Calibration Is Used.
Excitation Error Can Be Ignored If Excitation Is Not Being Used.
Implicit To The Error Specification Is The Requirement That Each
Channel Has An Individual Regulation Circuit.
Without Individual Regulation Excitation Error Can Become Huge
Due To Shorted Or Open Transducers Affecting Each Other.
Site
Specifics
Measurement Accuracy: Real Or Imagined
Group Regulation
Individual Regulation
Red Transducer Wires Short
Out
Red Transducer Wires Short
Out
Do Not Fall Prey To This
Transducer
Output
Transducer
Output
Measurement Accuracy: Real Or Imagined
Signal Conditioning Error
Signal Conditioning Errors Affect Bridge Measurements.
The Balance Error Specification Can Be Ignored If End To End EU Calibration Is Used.
Signal Conditioning Includes Bridge Completion, Bridge Balance, & Possibly Others.
Different Assumptions Will Lead To Different RSS Numbers
Assumptions Depend On Your Testing Requirements.
Site
Specifics
Measurement Accuracy: Real Or Imagined
Amplification Errors Are Worst At Highest Gain
The Zero Accuracy & Linearity Error Specifications Can Be Reduced By Using
End To End EU Calibration Or Voltage Calibration.
CMRR Is Normalized Into A Value Using CMRR = 20 log10( Gain/CM Gain) And
Vout = CMGain * (Voltage Measured). Example Assumes
Measurement Is FS Input.
Assumptions Depend On Your Testing Requirements.
Amplification & Filtering Error Site
Specifics
Measurement Accuracy: Real Or Imagined
Digitization Error 16 Bit SAR
Reference IMAC 2008 Presentation
16 Bit SAR Used With High Gain (5000) Provides 9781 Effective Codes
Per mV Measured.
This Error Collection Is Normalized To 2 mV Full Scale Measurement &
Matches The Normalization Used In Determining Amplification & Filtering
Error.
Site
Specifics
Measurement Accuracy: Real Or Imagined
Digitization Error 24 Bit
Reference IMAC 2008 Presentation
24 Bit Has Much Lower ENOB Due To SINAD Numbers For Conversion
Technique.
This Error Collection Is Normalized To 2 mV So That The Higher Gain
Product (SAR) Can Be Properly Compared.
Coupled With Lower Gain (Typical Of Offerings) Leads To Worse
Error Numbers When Considering Level To Measure.
Site
Specifics
Measurement Accuracy: Real Or Imagined
+
Always Keep This In Mind
No Single Element Of A Measurement System Will Optimize Measurements.
All Elements Play A Part In Measurement System Accuracy.
Element Worth Depends On What You Intend To Measure And
How You Intend To Use The Element.
Measurement Accuracy: Real Or Imagined
Cabling Error
Every Connection Point Along The Path From Amplifier To Transducer
Creates An Electromotive Force (EMF).
EMFs Add Errors. These Errors Can Be Due To Plug On Modules The Measurement
System Uses To Allow Customization Of A Standard Input Module And Save $$.
Source Currents In The Amplifiers Cause Errors Also When They Work Against
Slight Mismatches In Resistances Along The Way To The Transducer.
Can Be As High As +/- 0.1 % At High Gain (Low Level Measurement 2 mV FS)
R1
&
R2
Switching
Source Currents
I1 & I2
Protection
R3 & R4
Connector
R5 & R6
&
EMFs
Cable
R7 & R8
&
EMFs?
(I1 I2)
(R1 R2), (R3 R4), (R5 R6), (R7 R8)
Influenced By You
Measurement Accuracy: Real Or Imagined
Summing Up Measurement System Errors
Assessment For 2 mV FS Measurement At High Gain
This Assumes All Errors Come Into Play (Not Generally The Case)
Calibration Brings The Error Down Significantly
Remember, The Transducer Was +/- 0.173 %
The Combined RSS (Calibrated) Is +/- 0.202 % Indicating The
Transducer Is A Strong Controlling Element Here.
Calibration Errors Do Need To Be Derived And Included.
Measurement Accuracy: Real Or Imagined
Voltage Calibration Issues To Watch
Switching
Source Currents
I1 & I2
Protection
R3 & R4
Connector
R5 & R6
&
EMFs
Cable
R7 & R8
&
EMFs?
(I1 I2)
(R1 R2), (R3 R4), (R5 R6), (R7 R8), (R9 R10)
R9
&
R10
Voltage Calibration Electrical Path Difference Manifests In Offset Errors
When Compared To Point A Or Point B.
Characterizing The Difference Between A and B Allows For Removal
Of All Source Current Affects.
This Typical Voltage Calibration Arrangement Will Not Lead To Great
Absolute Level Indications If A and B Are Not Characterized.
It Is, However, Very Useful In Eliminating Slope (Gain) Errors.
A
R1
&
R2
B
Measurement Accuracy: Real Or Imagined
Trusting Calibration
Calibration Sources Must Be Calibrated By Metrology Labs
Calibration Data Must Be Checked For
Hysteresis Errors
Noise Errors
Linearity Errors
Flat Line Errors
Overload Errors
80%
60%
40%
20%
0%
-20%
-40%
-60%
-80%
80 60 40 20 0 -20 -40 -60 -80
Percentage of Full Scale EDC Input Voltage
Percentage
of Full Scale
Measured
By Channel
Conversion Equation
Voltage Step Mean
Measurement Accuracy: Real Or Imagined
Trusting Specifications
Calibration and Test Data Reports
Detailed report including pass/fail criteria
Gain accuracy (all outputs)
Linearity (all outputs)
Zero accuracy (all outputs)
Bandwidth (all outputs)
Filter response
Noise (all outputs)
Common mode (350 Ohm)
Source current
Cal attenuator accuracy
Current exc. accuracy & regulation
Voltage exc. accuracy & regulation
Voltage monitor accuracy
Bridge completion function
Autobalance accuracy
Shunt calibration accuracy
Final System Test Report
You
Should Ask
For This
Information
Measurement Accuracy: Real Or Imagined
Concluding Remarks
No Single Element Of A Measurement System Will Optimize Measurements.
You Need To Understand How The Specifications Affect Your Testing.
The Transducer Will Have A Large Affect On Measurement Accuracy
Since Its Environment Is Not Generally Controlled.
Measurement System Environment Needs To Be Taken Into Account.
If A Specification Is Missing Or Obtuse Ask For Clarification.
If The System Performs Better Than Specifications Indicate Then
You Have Found A Good Vendor.
If The System Performs Worse Than Expected Either You Are Not Using
It Correctly, You Have Interpreted The Specifications Wrongly, Or You
Are A Victim Of Specsmanship.
Good Software Can Help, But It Does Not Make Up For
Poor Transducers Or Poor Measurement System Hardware.